1.6 KiB
1.6 KiB
Example Workflow
This repository includes a sanitized example workflow:
workflows/z-image_5.4_emberframe.json
The workflow demonstrates a Z-Image Base / Z-Image Turbo split-sigma sampling chain using EmberFrame helper nodes and PiD final-latent decode.
What Was Sanitized
- private/custom LoRA filenames were removed and LoRA slots are disabled
- custom merged checkpoint names were replaced with placeholders
- local image inputs were replaced with
example_input.png - temporary preview image metadata was removed
- prompt-helper notes were replaced with public-safe example text
- old private wildcard references were replaced with the bundled example wildcard files
Before Running
After loading the workflow in ComfyUI, set these to your local files:
- Z-Image Base diffusion model
- Z-Image Turbo diffusion model
- VAE
- CLIP/text encoder
- PiD checkpoints
- optional ControlNet, depth, upscaler, and SeedVR2 models if you use those sections
If you use the img2img path, replace the placeholder example_input.png with your own image.
Required External Node Packs
The workflow still needs external packs such as:
ComfyUI-PiDrgthree-comfyComfyUI-Image-SaverComfyUI-Easy-UseComfyUI-KJNodesRES4LYF- ControlNet/depth helper nodes used by the optional sections
- SeedVR2 / Ultimate SD Upscale packs used by the optional upscaling sections
Install emberframe-nodes once for the EmberFrame helper nodes used by the main sampling and prompt-helper sections.
PiD Reminder
For Z-Image / Flux PiD, keep PiD scale at:
scale = 4
Both the 2k and 2kto4k PiD checkpoints are 4x decoders.